![]() |
![]() |
#1 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
'AmiLFS' system build ~ inspired by AmiDeb
#AmiLFS is free unlicensed text files ~ use any way you see fit, doin' the funky gibbon is encouraged =)
#Status ~ WIP #Location ~ https://github.com/giantclambake/AmiLFS #Target Machine ~ 2ndhand 'Thin Client' ix86 based, intel i915 video chipsets (see my next post) #Deployment type ~ initially live_boot USB stick ; later, this + install to hdrive #Intent ~ testing only, provide clear instructions or how to recreate deployment stage And now, the TLDR; Recently I was looking at the AmiDeb installation, seeing if I could leverage some old ix86 machines here, or then with a curiosity to see just how big an ix86 monster you needed, to emulate an m68k Amiga @ 7Mhz using the AmiDeb linux instance. I likely achieved that goal. Somewhere around here is a TLDR post that's been moved and I cannay find it, but, to quote self here, I posted this para; "In this regard, I think I can do better than Debian (or Arch for that matter), but that is conjecture ~ I have to know, for my own edification, whether or not that conjecture holds water. Thus....I'm going to toddle off, dump debian base, and build AmiDeb into a proper linux appliance. I'll keep notes as I go =) It's been nigh on 10 and a bit years since I mucked about with linux builds...but it's ok -- I still wake up in the middle of the night screaming "configure, make, make install"...it's like learning to ride a bicycle ~ once learnt, the skill tends to never leave you...building a linux system is the same 8)" I see now, I failed to 'under-write' that properly, in that "I think I can do better than Debian (or Arch for that matter)" appears as a point of possible boasting or self-bravado, largely because it's so far removed from context ~ mia culpa ~ the correct way to express that, would be 'anyone who's done, or chooses to do, an LFS build of the AmiDeb construct (using LFS instead of Debian), might feel/find they can do a more custom, specific linux build, that may actually run better/faster on the same ix86 hardware, than using an existing linux distro to do the heavy lifting'....and the proviso there, is 'if building to a specific ix86 target machine'. There is only 1 way to find out 8) In that statement I am affirming that AmiDeb will *always* be more user friendly wrt installation, more easily/reliably upgradable, and more generally applicable to the widest range of ix86 machines out there (even imacs...muhaha =) that users of the world would use AmiDeb on...off into an unknown future, as it rides on the shoulders of that giant we call Debian. In contrast, the LFS based build I'm working on would not be an installation like AmiDeb/Debian, but more a 'deployment' image, which isn't really user friendly, and whilst being upgradable, you need to have the same linux knowledge/skills to recompile any upgraded software, and remake the deployment image and over-write/replace what's required, and that image is only going to work on a specific criteria of of ix86 setups (1 sound driver, 1 X driver, only kernel supports for that actual hardware you have/need..blablabla). 'AmiLFS' is a 'static' linux instance, ultimately speaking ~ that is, it is based/built upon a specific version of the LFS book, using specific tarballs of the required software packages, and following specific pages in the LFS & BLFS books. Eventually, the system software build tree may figuratively 'break' ...ie; if some feature or performance improvement of X or fs-uae, ends up relying on newer versions of gcc/glib/libc, you pretty much know you might have to go back and start at the beginning again...even back to toolchain. I know this may not be, everyone's cup of tea, but I did say I would keep notes, and that's ongoing....set github above End of the day, right, it's non trivial, time consuming, to build your own linux system, so I'm making build images available to -save- ppl's time, should they want to do it for themselves and be able to jump into the LFS/BLFS books at key moments in the build process, or just wait for the deployment image(s) when it's all done & dusted. This is important, for wrt peer review ~ if it does run better/faster, others must be able to recreate the same results, on -their- hardware, and not just take my word for it. Yes, it feels kinda strange saying that, when I don't yet really know for sure, if AmiLFS is any faster than AmiDeb or not =) footnote: I started hobby interest in the late 70's with the advent of the z80, then ended up doing industrial process computer control on archaic mainframes coded/running BCPL, and when leaving that field found I finally had time for Personal Computing, and in the mid 80's that was a no brainer - Amiga. Obviously we all saw what happened by the mid 90's to the Amiga platform, and from there I went to linux on ix86 ...still here =) |
![]() |
![]() |
#2 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
#Target Machine
The actual build rig is amd64 CPU & nvidia GPU so I aim for that test target ; it will work on Thin Clients with nvidia GPU be it in a slot or onboard. If you post your exact Thin Client make/model, I can divine whether it'll work or not. In the wild, a good many of these Thin Client boxen, have one or another x86_64 CPU, and intel IGP for graphics, that's catered for by the Xorg i915 driver. Plan is to get one of these wee beasties, and due to the driver diversion @ mesalib build specifically to that target (not nouveau/nvidia), because as I say most of the cheaper units need this sort of treatment. If you want to run/test AmiLFS on a Thin Client that uses AMD APU or onboard ati/amd GPU chipset, you'll have to let me know (or do it yourself =) l8r |
![]() |
![]() |
#3 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
Build to fs-uae target completed just before midnight, Fri 10 Feb 2023 AEST
Configurata begins on the morrow =) |
![]() |
![]() |
#4 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
Thin Client machines;
I spent a bit of time today looking at the 2ndhand market for these machines (read: ebay) It would seem one of the reasons behind the why of there being so many of these devices around at the now, is thanks to the TPM requirement introduced by win11 (a lot of these 2ndhand units have win10 installed, activated)... AMD GX-424CC SOC with Radeon R5E Graphics AMD Ryzen Embedded R1505G with Radeon Vega Intel Core i5-7200U CPU @ 2.50GHz (IGP graphics core depends on exact model, but they all use the linux i915 driver) ...anything near the above with 4~8GB ram is something near the ballpark (intel Atom, probably not). The intel NUC type mini systems would also be a good target, anything really small with 1 x pcie slot (for discrete GPU) would be ideal. You likely want to avoid anything that has a soldered-2-board eMMC chip for internal storage (and no M.2/nvme/sata 2nd drive support)...we all know why.. (looking @ u Apple 8) .. As I did say my first 'real' target machine was going to be i5/i7 based with i915 supported IGP, and dragging in the fact I was to firstly going to deploy things as a USB bootable 'live' instance, having USB3 ports has appeal...ergo.. https://i.dell.com/sites/csdocuments...Spec-Sheet.pdf I found 1 on ebay for ~AUD$55 ...with 8gb ram/120gb ssd. This seems like a very common/plentiful thin client spec of i5/i7 + intel HD4xxx IGP. It'd be nice to find a All-in-One PC Thin Client with enough grunt, but that doesn't seem plentiful and/or they're underpowered (perhaps stick a rpi in one, or find one based on ARM where you're not on ix86 arch.. ...this is the first intel CPU based system I've bought since '94 ... B) |
![]() |
![]() |
#5 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
M'kay... if you're following this text, you should have correctly concluded the fact that me going out and buying specific target ix86 hardware, infers that I saw 'something' while testing AmiLFS over the weekend just gone, that made me consider this is worth looking into further. On the build machine, I'm sortta at that spot where I have to figure out what's needed to get the original nvidia 304.108 tarball to compile =) This means any sort of comparative testing is at the moment limited to nouveau drivers, and I have other things that make 'comparative' testing nigh on impossible to achieve...
....first place I'm looking for speed, is at initial boot/start of the underlying linux system itself. To truly test that, requires all systems to be on the same hdrive technology...and that is very much not the case right now, with the debian 11.6 build host on a ST 1.5TB sata spinning disk, AmiLFS on an even older WD 160GB sata spinning disk...and AmiDeb on a 64GB Fujitsu sata SSD =) The Optiplex-9020 has a M.2 slot and the OS image will end up there eventually, but it does have 2 x sata ports, thus a direct comparo between AmiDeb & AmiLFS is possible. Thus I've ordered 5 x 64GB Fujitsu SSDs (AUD$11 each ... nobody wants them =) to really level the playing field. AmiLFS easily beats Debian 11.6 to login prompt (single user console mode), but impossible to call against AmiDeb due to the SSD advantage it has... ....there will be some 'glass ceiling' here somewhere, where the Amiga emulation engine can only run as fast/accurately as it can, and probably you most see this wrt the amount/frequency of dropped frames. Underneath this, is how well the linux kernel is able to run the emulator, and get the data to the display. At this base level, there should be something to see in linux benchmarks ~ a more simplistic test for example, would be something like 'startx /usr/bin/glxgears --fullscreen' (as root) from commandline (you need be sure X hits the display res it presents defaultly in all cases). I already know wrt the build/test machine, glxgears is going to be better off with nvidia drivers, but on the Optiplex-9020 there will be no such advantage to either party...but it is something I'll test later... ...as for AmiLFS itself running fs-uae, it is a little bit better ~ that scrolling text I point to above in the Andromeda.DOS demo is visually smoother for example...I'll need setup the vidcam on tripod and do some rudimentary time measurements (times like this one wishes for a highspeed camera =). AmiLFS is currently running linux-6.1.7-RT If you'd like to listen to some of the (no longer widely available) music I listen to doing AmiLFS , check the github page ...more l8r... Last edited by giantclam; 18 February 2023 at 12:33. |
![]() |
![]() |
#6 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
M'kay..... glxgears tests done, but I should note the differences and test goals...
AmiDeb is close to Debian 11.6, so software version are nearly the same ~ that said, the AmiDeb installation (x86) here is using a RT kernel, but the Debian instance is on stock release kernel.... AmiLFS however, is closer to Debian sid and you see this reflected in software versioning ~ AmiLFS is also running a RT kernel, but a much newer release version. The one thing constant here, is the hardware ; it was operating in 25C air conditioned comfort wrt all tests. *The stock Debian instance needs a 'systemctl set-default multi-user.target' issued & reboot, to get it to boot into the same runlevel as AmiDeb *The menu.sh script that's part of AmiDeb, issues a 'sudo cpufreq-set -g performance' command ~ one needs to do the same thing on Debian 11.6 to obtain the same kernel performance level (the AmiLFS kernel builds have this option compiled in by default, so no need to use cpufreq-set command) *Both Debian 11.6 & AmiDeb do not include the glxgears binary, so it needs to be added -- sudo apt install mesa-utils Wrt running the fs-uae emulator, what you're really looking for here is any 'jitter' ~ this is reflected in the time wallops, not the FPS raw value (301 frames in 5secs)...but rather the FPS value down to milliseconds -- the flatter the time line, the better. Starting with Debian 11.6, prepared as above.... Command used for all tests, issue by root: startx /usr/bin/glxgears -fullscreen &>run.log Debian 11.6 x86 ; X.Org X Server 1.20.11 X Protocol Version 11, Revision 0 Build Operating System: linux Debian Current Operating System: Linux debian11 5.10.0-21-686-pae #1 SMP Debian 5.10.162-1 (2023-01-21) i686 Kernel command line: BOOT_IMAGE=/boot/vmlinuz-5.10.0-21-686-pae root=UUID=7357a9e6-865b-439c-bde1-068d7782b12e ro quiet Build Date: 17 December 2022 10:00:08AM xorg-server 2:1.20.11-1+deb11u4 (https://www.debian.org/support) Current version of pixman: 0.40.0 Before reporting problems, check http://wiki.x.org to make sure that you have the latest version. Markers: (--) probed, (**) from config file, (==) default setting, (++) from command line, (!!) notice, (II) informational, (WW) warning, (EE) error, (NI) not implemented, (??) unknown. (==) Log file: "/var/log/Xorg.0.log", Time: Mon Feb 13 18:37:39 2023 (==) Using system config directory "/usr/share/X11/xorg.conf.d" (II) modeset(0): Initializing kms color map for depth 24, 8 bpc. Running synchronized to the vertical refresh. The framerate should be approximately the same as the monitor refresh rate. 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.021 FPS 301 frames in 5.0 seconds = 60.021 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.004 FPS 301 frames in 5.0 seconds = 60.007 FPS 301 frames in 5.0 seconds = 60.009 FPS 301 frames in 5.0 seconds = 60.011 FPS 301 frames in 5.0 seconds = 60.012 FPS 301 frames in 5.0 seconds = 60.013 FPS 301 frames in 5.0 seconds = 60.014 FPS 301 frames in 5.0 seconds = 60.015 FPS 301 frames in 5.0 seconds = 60.016 FPS 301 frames in 5.0 seconds = 60.017 FPS 301 frames in 5.0 seconds = 60.018 FPS 301 frames in 5.0 seconds = 60.018 FPS 301 frames in 5.0 seconds = 60.018 FPS 301 frames in 5.0 seconds = 60.018 FPS 301 frames in 5.0 seconds = 60.019 FPS AmiDeb x86; X.Org X Server 1.20.11 X Protocol Version 11, Revision 0 Build Operating System: linux Debian Current Operating System: Linux AmiDeb-x86 5.10.0-21-rt-686-pae #1 SMP PREEMPT_RT Debian 5.10.162-1 (2023-01-21) i686 Kernel command line: BOOT_IMAGE=/boot/vmlinuz-5.10.0-21-rt-686-pae root=UUID=d00a44b2-30a7-4117-b116-276724783223 ro quiet splash Build Date: 01 February 2023 02:11:18PM xorg-server 2:1.20.11-1+deb11u5 (https://www.debian.org/support) Current version of pixman: 0.40.0 Before reporting problems, check http://wiki.x.org to make sure that you have the latest version. Markers: (--) probed, (**) from config file, (==) default setting, (++) from command line, (!!) notice, (II) informational, (WW) warning, (EE) error, (NI) not implemented, (??) unknown. (==) Log file: "/var/log/Xorg.0.log", Time: Mon Feb 13 18:57:51 2023 (==) Using system config directory "/usr/share/X11/xorg.conf.d" (II) modeset(0): Initializing kms color map for depth 24, 8 bpc. Running synchronized to the vertical refresh. The framerate should be approximately the same as the monitor refresh rate. 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.024 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.024 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.024 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.025 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.024 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.023 FPS AmiLFS x86 ; X.Org X Server 1.21.1.4 X Protocol Version 11, Revision 0 Current Operating System: Linux <amilfs> 6.1.7-rt5 #1 SMP PREEMPT_RT Sun Feb 12 17:03:27 AEST 2023 i686 Kernel command line: BOOT_IMAGE=/boot/vmlinuz-RT-NV-MCP61 root=PARTUUID=85c9b003-06 ro Current version of pixman: 0.40.0 Before reporting problems, check http://wiki.x.org to make sure that you have the latest version. Markers: (--) probed, (**) from config file, (==) default setting, (++) from command line, (!!) notice, (II) informational, (WW) warning, (EE) error, (NI) not implemented, (??) unknown. (==) Log file: "/var/log/Xorg.0.log", Time: Mon Feb 13 19:11:23 2023 (==) Using system config directory "/usr/share/X11/xorg.conf.d" resize called 1280 1024 Running synchronized to the vertical refresh. The framerate should be approximately the same as the monitor refresh rate. 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.022 FPS The Debian 11.6 is all over the shop ; there's nothing wrong here, but a full debian system loads/runs a lot of services ; if you matched the init path to AmiDeb, you'll like see less variance. AmiDeb does lots better, as it's not loading/running everything Debian 11.6 is, but it is running a debian patched kernel....(not one specifically compiled to suit the specific hardware used) AmiLFS wins here, with a deviance of +/- 1ms ... but I'd be 'reasonably confident' AmiDeb might achieve better/similar results, with the same kernel version (pristine, unpatched) compiled to the hardware it's running on. We also gain a little bit in boot time (grub -> login console), as AmiLFS doesn't twiddle it's thumbs loading an initrd image. Dumping to logfile also steals cpu cycles, and as said, because all instances are on different disk types, there's likely some variance there as well --- when all 3 instances are on the same SSD type, I'll redo these tests to discover just how much disk access has to do with variances observed here...BUT...AmiLFS is on the oldest, slowest spinning disk, and we're still close to flatline.... ie; if disk access was participle here, it should become obvious =) ...more l8r... footnote: yep, I put versioning in kernel =) |
![]() |
![]() |
#7 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
[Teaser]
What is the default control device used by game consoles to navigate the menu? This is what I'm aiming for wrt AmiLFS menu navigation... ![]() |
![]() |
![]() |
#8 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
I did a hand-rolled spliff of AmiDeb ontop of debian testing (bookworm/testing), primarily to provide an easy path to nvidia old legacy drivers (340.xxx) in debian unstable (sid)...
...it was around here I found (using apt) that the fs-uae-launcher package was not in the (testing/unstable) debian repos, and so I needed to backtrack that target to find out why (there is always good reason why something isn't in deb repos). Turns out, last time that package appeared alive was back @ debian buster, at a time when the fs-uae tarball included fs-uae + arcade + launcher, but since then the fs-uae-launcher was moved out of this tree, and became a separate target. AFAICT, sometime after this juncture a packager's notice pointed to the fact, that the fs-uae-launcher was in a state of dimensional flux, and to hold off using it for packages (https://github.com/FrodeSolheim/fs-u...ster/README.md)... ....if one uses the FrodeSolheim repo, a similar situation prevails, and this is likely why fs-uae-launcher notes to screen 'expected fs-uae version <bla.woohoo!> but got version <bla.blabla> instead...the two trees are just a little bit out of step, I think just because above readme file.... fs-uae + arcade moves forward ; fs-uae-launcher is in a state of flux. Now...this all makes sense to me, as someone who's doing the LFS thing --> I defined my target to be fs-uae, blithely unaware the fs-uae-launcher had become a separate target of and unto itself. Did fs-uae + arcade build from git-master on the AmiLFS system? Yep, absolutely fine, haven't thrown any optimizations at it first time up, but default compiler flags should always be checked first imho ; I'll rebuild it later with a more rabid cflags... target achieved tho' [tick] ....can the AmiLFS system build the fs-uae-launcher target? Nope, not yet...apart from some python3 stuff that needs to be added, my minimum target is python3-pyqt5 which means I have to build qt5 and satisfying it's dependencies as I go (there might be some functionality I can leave out wrt what fs-uae-launcher requires) first, before I can hit the target of what AmiDeb already does. Having said that, the old 160GB spinning drive that instance is on, is running really hot, so I'm going to image it onto an sdd before going further... //test while testing ~ I got this ebay-plus email last week, shilling 'Netac' SSDs (made in CN 8)...120GB for AUD$15...this, I gotta see...try... it arrived today ; I'll image the build drive, and do a restore to this Netac and see if gcc & friends can flog it to death...lol.. ...more l8r.. |
![]() |
![]() |
#9 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
The thin client box gets here tomorrow ~ first off the rank, is to install and/or just boot to the AmiDeb x86 instance, to give some feedback to vk3heg wrt how that goes on that hardware IRL...
//...one day later...// It's here ~ using the well known layperson's measurement system of a RedHead matchbox, it's 1.2matchbox H x 4.5matchbox L x W.. ![]() Haven't started it yet, as it claims to be 'professionally refurbished' ~ this usually signifies 'due diligence' is required, to at least check the heatsink compound between CPU & cooler has been replaced....heed this advice ; these are hinky little machines, put together with the fold flap A under flap B after removing flap C so that flap D is released type constructions, and refurbishing them takes time and time costs money and this didn't cost too much of that. I'll be taking pictures along the way ~ if I find any evidence of it not having been 'professionally refurbished', seller's going to get an earful wrt 'item not as described'...(often worth a discount =) As I somewhat expected, hitting the Qt5/PyQt5 target, was akin to chasing a hare with a shotgun through a paddock of lantana 8) The LFS book offers comfort to the newcomer, however...like Klingon software, BLFS doesn't really coddle the weak ; especially if there's no target, and you need divine a way to the target yourself. I'm not ever sure what happened tbh...that is, my bash_history & timestamps on installed software and Makefile creation times, form the basis of my own 'package management strategy'...I'll review it later, as I need to figure out the automated build scripts... ....I now full well appreciate ?why? the fs-uae-launcher project is not being widely packaged & distributed as per it's git readme.md ... it's a minefield at the moment. PyQt5-5.15.0 was the only version I could get to build cleanly against qt-everywhere-src-5.15.5 ..which was weird, and the exact version of sip seems to matter... (and the qt5 configure switch '-pcre' was no help when it refers to libpcre2 =) Everything else via pipenv was more or less ok (using 'python3 -m pip <option> [target]'), and what didn't, built from source tarballs fine. Using the master branches of fs-uae/fs-uae-launcher from github now compile, being BASEVERSION_FS=4.0.40 & BASEVERSION_FS=4.0.99 respectively. I've already checked fs-uae + arcade which seem to work as expected & well, but it appears the launcher Makefile has been stripped of it's install target, so I'll have to set about manually doing that ; once there, I can edit the AmiDeb menu.sh file to account for my system makeup, and reach the 'inspirational' goal of flattering AmiDeb by way of imitation B^) footnote: just opened the thin client ; wrong screws forced into ssd mounting holes, CPU/heatsink interface not serviced...typical. Whenever I see 'professionally refurbished', it often ends up being visions of... ![]() ....I should add, this'll be interesting ~ I'm not sure if the boofle with the official fs-uae repo has been fixed, but I do know the thin client has win10 pre-installed, and this'll be time to engage 'windows dummy' mode, and see how easily/readily the current x86_64 AmiDeb install set blows windows away... ...more l8r... |
![]() |
![]() |
#10 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
target machine: Dell Optiplex 9020 USFF Desktop PC @2.90 8GB DDR3 120GB SSD
Thing is well built ; if you're lucky the unit has only done 9 to 5 work, in an air-conditioned highrise building, and lived the good life as far as these little machines go...actually, because it is Dell bios, it may have a uptime log in bios maintenance mode, so I can see how many hours the board has been on, cold booted, or whatever they want to log. It seems really nippy with debian 11 on it, but in another post I note AmiDeb failed to install to it :^/ Which brings up the question ~ how do I migrate the AmiLFS build, to the thin client box? Short answer is, I can't.. at least not the 'full test build' just completed, because 1> I'd need recompile the kernel to that hardware && 2> everything up to mesalib per-amble would be constant across all build targets, but everything around the mesalib build is different dependant on GPU type... ...you tend not to know these things going in, but march forward to test if you can get to the target(s) as thiftily as possible, and the first AmiLFS instance is very much hypnotized by the nvidia toad. Along that path as well, I made a ballsup somewhere wrt if I was ever going to build Qt5 or not, which I wasn't at the time, so I need redress that... ...the AmiLFS build up to compile/install kernel, install grub ..ie; $base ...is valid for the optiplex from that staging point forward. Likewise the debian x86 build host should boot & run on the dell hardware, but life's never meant to be easy, as although there's a 2ndary sata channel for the optical drive, it's been dell'd due to the sata power connect being 5v only, and there's only one full sata power connector for the ssd...and I'm dubious about loading that dinky little PSU up, with the suck of the old ST 1.5GB spinning disk -- it's easy to inadvertantly damage these units if you don't realize they're limits ..it likely tops out <100w ... the PSU has a green LED & small back plastic push button on top ; no second guesses this'll be the PSU current breaker, to cope with office situations of user pluging in every maximum power draw USB 2/3 devices they can while playing cod on worktime with their m$-office spreadsheet minimized...lol.. ...once I get that drive attached, I'm pretty sure the DellUEFI bios freaks out when you boot it up, in as much as it doesn't present the drive as a valid boot target...iirc, that means you have to cold boot with the bios maintenance jumper installed, and it sits there and twiddles it's bits before announcing 'you may now turn off the machine, and set bios jumper to normal position'...or such'n'similar...see how it goes. I'll have to steal power from the build machine, but should be all good... ...after booting the build debian instance, I can adjust my $LFS mountpoint variable, to point to the ssd in the thin client...then I can untar the base build tree, chroot to the ssd, configure/build a kernel just to suit that hardware, install it, install grub...and do the $BLFS thang all over again, but this time with tested maps that are know to work wrt reaching the fs-uae + arcade + launcher target... ...seeing as the start of Xorg (which hardware driver) build is the beginning, it's possible to re-order the build order, so that everything that -can- be compiled/installed before X that doesn't have a hardware diversion gets done first ; this would be the end of stage 2 ... sounds like waffle right now, but consider you've got one specific set of source archives that are never going to change, and you're building to an ix86 'platform' that likewise will never change, so you can automate this process and launch single commands to reach stage completions ..ie; stage 0 is building the initial toolchain. Hint: start here if you wanna think about cross-compiling for another -march/-cpu ... stage 1 is 1 command to get to $base system, stage 2 is 1 command to get to X/mesa/and friends, and then there's a choice of 3 automated build scripts, to hit either the AMD/radeon, nvidia, or intel target (there's actually more wrt laptop chipsets)... ...thus, anyone from the now and off into an unknown future, can rebuild the appliance from source, all self contained. It's 'timeless' in as much as the source archive selection is static, down to the selection of kernel. You never plan to upgrade anything, as long as x86/x86_64 is alive and kicking, this will build...in theory. At some stage x86/x86_64 will becomes something this lot won't build on ..not holding my breath on that one just yet ; lots of this hardware out there... =) ....lol....and to remark to the more than coincidental, the ssd unit is a 128GB Netac brand drive...what are the odds? ; I'll interrogate it later and see what it's boot/life figures look like... Hmmm...I know what I should do next -- plug the ssd with AmiDeb already installed into it, and see what happens 8) |
![]() |
![]() |
#11 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
M'kay, transmogrification of the $base system to the optiplex done...
![]() Just while it's at this point (stage 1 ~ base system only, quick boots =), I've been fathoming a path to get it to boot 'how I want it to' -- that is, I want it to boot like a games console, not a linux system ... you know, switch on, machine boots to (animated) splashscreen, which disappears leaving the main user UI visible. To do this as 'seamlessly' as possible, you need to select (and keep) the same display resolution as the user UI is going to use, a best way to do that is not use a bootloader, and instead go EFISTUB in kernel, set the efifb/fbdev when there, and during init make sure vga_modeset is set accordingly ~ this won't interfere with X, but the i915 fbdev is a strange thing I'll need look into a bit better (it might be easier to use simple_vga console instead)... ...created a (U)EFI/legacy bios, USB bootdrive ... this is sort of in preparation from the deployment stage -- I also looked at that 'hidden OS' aspect, wherein machine boots to emulated Amiga gamebox by default (and there's no way for userland to stop that), but if you plugin the boot/service USB stick, machine boots up debian, or LFS on minimalist TWM/console..so that all works. On an Optiplex 9020 anyway, which is all it has to work on (it'd work on other similar hardware). In a lot of ways, this is like working on Amiga (or any other 'fixed' hw platform) -- there are no slots to plug something else into, all the sound/video hardware remains the same...and although one could put a different CPU in it, you're limited by TDP (and what the PSU can give), and it's still going to be covered by the i915 makeup and doesn't matter. There will (and has) always been debate about the relative merits of a highly modularized kernel build vs something with the junk all builtin on the inside. Long ago I did that test (back at linux-2.x.x), and for a DAW at least, there seemed little difference between the two --- now, some 20years later, I should do it again..just to sate my curiosity.. root [ ~ ]# lsmod Module Size Used by mei_pxp 16384 0 snd_hda_codec_realtek 135168 1 snd_hda_codec_generic 69632 1 snd_hda_codec_realtek ledtrig_audio 16384 1 snd_hda_codec_generic snd_hda_codec_hdmi 53248 1 x86_pkg_temp_thermal 20480 0 snd_hda_intel 36864 0 kvm_intel 286720 0 input_leds 16384 0 snd_intel_dspcfg 16384 1 snd_hda_intel joydev 20480 0 snd_hda_codec 118784 4 snd_hda_intel,snd_hda_codec_hdmi,snd_hda_codec_generic,snd_hda_codec_realtek kvm 802816 1 kvm_intel snd_hwdep 16384 1 snd_hda_codec irqbypass 16384 1 kvm led_class 16384 2 input_leds,snd_hda_codec_generic snd_hda_core 81920 5 snd_hda_intel,snd_hda_codec,snd_hda_codec_hdmi,snd_hda_codec_generic,snd_hda_codec_realtek snd_pcm 110592 4 snd_hda_intel,snd_hda_codec,snd_hda_core,snd_hda_codec_hdmi snd_timer 32768 1 snd_pcm mei_me 36864 1 snd 73728 8 snd_hda_intel,snd_hwdep,snd_hda_codec,snd_timer,snd_hda_codec_hdmi,snd_hda_codec_generic,snd_hda_codec_realtek,snd_pcm mei 81920 3 mei_pxp,mei_me ..with no X yet. I've left the sound stuffs out as modules at the now...and I've got no idea who would use DisplayPort to HDMI, but it's there 'coz teevee displays and cables for this exist...I think? =) The little optiplex 9020 is a wee champ btw, compiles lots faster than the Phenom X4 in the box beside it (above). I just gotta transfer some archives to it, put it all back together.... then rebuild the system (not in chroot) ...this time, to the i915 target and aim for fs-uae-launcher ...and keep md5sums & notes for the automated alfs scripts....and I should grab the alfs scripts for that LFS book revision, and adjust/edit/test them to rollin the errata + extra archives to build before boot... ....the $base system {ahem} 'installation' (...ie; 'cd $LFS && tar -xpf $HOME/amilfs-base.tar.xz' ... then mount filesys and chroot, configure/install kernel, edit networkconf, reboot 8)...works, and passed gcc and other tests once booted, means the original amilfs system has earned the name 'pilot' (FarScape =)...the instance on the optiplex 9020 I haven't decided a name for yet... ...I know what a good bootsplash screen would be though...the usual Tux the penguin image with either an Amiga bouncing ball anim..or floppy disk load anim... something that says 'Amiga' in the same way Tux says 'linux' (there's a job for someone else not me =)... ...more l8r... |
![]() |
![]() |
#12 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
// IRL ... I was up to around 03:00hrs AEST building the i915 system, and when I finally reached LLVM, knowing it takes some time to compile, I just set the build going, and went to bed..hopeful that when I awoke, I'd find a clean build. Well, that didn't happen as planned ~ the coppers raided the house/property this morning ... lol ... startled awake by loud banging & shouts of "Police!", me dazed and confused trying to get to grips with the wtf was happening. Eventually, an hour later, they determined I had nothing to do with the warrant, and I jumped in the car to 'leave the scene area' as it were..'coz they wouldn't let me back in the house, until they'd finished... bugger. When I got back, the cops had left, and I just looked at my house-mate and said "Well, I didn't expect the spanish inquisition for breakfast.." ...whooosh!... 8) I should be fair, and say these coppers (QPS) were a 'good bunch of blokes' in aussie venacular, did their job well...a far cry from back in the 70's and what was the police state of queensland under the rule of JBP and Big Russ ... 'the minister for everything' .. lol, fun days... =)
Back @ the splatterplank, LLVM did compile fine with a parallel version matching clang, and wasn't the foobar it was on the nvidia build. It could be just a consequence of me re-ordering the build order, or, something in LLVM is comparing -march=i686 to what it finds otherwise (x86_64) and making the wrong call (this might explain why on the nvidia build, I had to specifically add the -march def to $CFLAGS)... ...frigging 'intel inside'...I've stuffed the kernel up somewhere wrt fbdev, and the last thing you see is vga-arb saying "G'day...I'm turning the console off..", and it blackscreens =)... root [ ~ ]# dmesg | grep vga [ 0.171508] pci 0000:00:02.0: vgaarb: setting as boot VGA device [ 0.171508] pci 0000:00:02.0: vgaarb: bridge control possible [ 0.171570] pci 0000:00:02.0: vgaarb: VGA device added: decodes=io+mem,owns=io+mem,locks=none [ 0.171654] vgaarb: loaded [ 0.266692] i915 0000:00:02.0: vgaarb: deactivate vga console [ 0.268392] i915 0000:00:02.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=io+mem ![]() It's still running/booting fine, just no visible console... think laptop display with no backlight, and we have the invisible login prompt. Others have shared in this experience on one or another distro, with their resolve being rolling back a kernel version ~ I'll go forward if anything, but point is it >should< work as configured, but it's likely I have to throw the kernel some bootparams via grub...or I stuffed up B) Hasn't been in the way until now, where I want to check X is ok...so fix that now ~ the actual kernel config is a build-map for this machine hardware, and it has to be 'right' more than not =) I think it might be time to lift it's lid, and set the bios maintenance jumper, if for no other reason than to see what's on offer.. ie; in normal mode there's no setting to show POST screen, so that'll likely be there...but fix the damn kernel first, then test X, and if all good head for the fs-uae-launcher target. Along the way I discovered there's some intel IGP test suite, and I think there's a debian package for it as well, and although we all know what intel get up to wrt benchmarking <grin>, it might be a useful comparo tool for later wrt an $LFS build versus Debian. ///...later.../// Service mode of bios, just opens up cloud computing (network) settings, nothing else exciting...no OC menu item B) ///...later still...// All done & build map tested -- see https://github.com/giantclambake/Ami...ex-rebuild.txt Now I just have to configure it to init like AmiDeb, and see how it goes.... I notice vk3heg has released another cut of AmiDeb, so I might just check that to see if Optiplex 9020 is a valid target (yet..fails on previous AmiDeb image)... ....and now for some ear&eye candy I uploaded waiting for pyqt5 to compile ... [ Show youtube player ] |
![]() |
![]() |
#13 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
[0.000000] DMI: Dell Inc. OptiPlex 9020/014GRG, BIOS A25 05/30/2019
Code:
[0.170890] pci 0000:00:02.0: vgaarb: setting as boot VGA device [0.170890] pci 0000:00:02.0: vgaarb: bridge control possible [0.170890] pci 0000:00:02.0: vgaarb: VGA device added: decodes=io+mem,owns=io+mem,locks=none [0.170913] vgaarb: loaded [2.339806] i915 0000:00:02.0: [drm] VT-d active for gfx access [2.339902] i915 0000:00:02.0: vgaarb: deactivate vga console [2.341181] Console: switching to colour dummy device 80x25 [2.341220] i915 0000:00:02.0: [drm] Transparent Hugepage support is recommended for optimal performance when IOMMU is enabled! [2.341226] i915 0000:00:02.0: [drm] DMAR active, disabling use of stolen memory [2.341927] i915 0000:00:02.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=io+mem:owns=io+mem [2.354838] [drm] Initialized i915 1.6.0 20201103 for 0000:00:02.0 on minor 0 [2.355850] ACPI: video: Video Device [GFX0] (multi-head: yes rom: no post: no) [2.355965] input: Video Bus as /devices/LNXSYSTM:00/LNXSYBUS:00/PNP0A08:00/LNXVIDEO:00/input/input6 [2.382506] fbcon: i915drmfb (fb0) is primary device [2.463376] Console: switching to colour frame buffer device 160x64 [2.483991] i915 0000:00:02.0: [drm] fb0: i915drmfb frame buffer device M'kay...missed an opt in kernel, so need to blend it one more time...but at least the console is now working, and the i915 fbdev is set correctly (damn, the things you don't consider...a "domestic" release of OptiPlex 9020 thin-client machine doesn't have DisplayPorts, and most of them don't have a M.2 slot)... ...speaking of which I'll have to get a M.2 drive for it... this opening it up to flop on another drive to test AmiDeb with is getting old... quick 8) root [ ~ ]# uname -rv 6.1.7 #4 SMP PREEMPT_DYNAMIC Mon Feb 20 02:24:22 AEST 2023 Hmmm...clock's wrong =) Must be set to localtime not UTC.... Code:
301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.017 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.017 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.017 FPS 301 frames in 5.0 seconds = 60.022 FPS Well ...that's not too bad wrt glxgears without tuning... +/- 3ms jitter or so ...hmm, wonder if fs-uae is ok... Code:
[FSE] [WIN] Using scale 1.00 [FSE] [WIN] Creating OpenGL context from main/video thread [FSE] [WIN] OpenGL vendor: Intel [FSE] [WIN] OpenGL renderer: Mesa Intel(R) HD Graphics 4600 (HSW GT2) [FSE] [WIN] OpenGL version: OpenGL ES 3.2 Mesa 22.1.7 [FSE] [WIN] SDL_GL_SetSwapInterval(0) [...] Init video module [...] Init frame module [FSE] [VID] No format flags, setting default format [FSE] [VID] Initializing OpenGL video renderer ...seems okay, but duh...I've screwed the fs-uae build over ....go back and fix that...but roll back too... ...I'll need get a DP->HDMI cable to check out that output, but I think it's fine to switch sound device order... ...time for another kernel, fix that htbl gaff, slim down a bit more... tweak the governor... ....system is only a hop skip jump from IceWM, and whilst AmiDeb uses Samba, I'm going to use ftpd...... ... all electrical computers furniture bedding & flooring ..... [ Show youtube player ] |
![]() |
![]() |
#14 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
M'kay, that's a lot better, and just a tiny bit faster than the Phenom X4...
...RT kernel will be next ... 6.1.7 #4 SMP PREEMPT_DYNAMIC Mon Feb 20 02:24:22 AEST 2023 Code:
301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS .... 6.2.0-rc3-rt1+ #1 SMP PREEMPT_RT Tue Feb 21 10:36:40 AEST 2023 .... Code:
301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.017 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.017 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.017 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.022 FPS 301 frames in 5.0 seconds = 60.016 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.016 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.023 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.016 FPS Code:
301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.019 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS 301 frames in 5.0 seconds = 60.020 FPS The soundcard issue (for me) is as simple as popping a snd_hda_intel.conf in /etc/modprobe.d containing the line 'options snd_hda_intel index=1,0 (debian do it differently via /etc/alsa/* scripts, to divine which card# has /dev/dsp & /dev/mixer ...this'll be a saner approach that suits multiple machine types, where the snd_hda_intel module options may not give consistant results across all HD-audio chipsets) ; I don't have to worry about this ~ I'm targeting one specific ix86 machine makeup...but if you were aiming for multiple machine types, the debian approach is better =) Trod on one of those mines I mentioned earlier ~ in trying to build/install fs-uae 3.1.xx , I discovered it relies on pysqlite3 ....which is a python2 module, and I've built to python3...which provides it's own sqlite3 module...which fs-uae 4.x.x uses of course, but not fs-uae 3.1.xx .. =) I'm not installing Samba/SMB stuff because I personally don't believe that's the 'lowest common denominator' among OS' ; I believe the lowest common denominator for file transfer is ftp protocol....I'm old, m'kay =) I'd be sure every OS has one or another FTP client (midnight commander has one builtin), and we all know the Amiga has a number of FTP clients...(I mean, some folks might want to upload files from their real Amiga into the emulator..n'est-ce pas? =), and likewise SAMBA isn't a default installed package for a lot of linux distros, and even if it is, the end user needs to configure it appropriately.... ...so what's the catch here? Answer is 'client discovery' ~ AmiDeb works around this, by displaying the machine's current IP at the top of the menu screen, however, my final goal here is not to have a menu, so I need another means. One could login to their modem/router, and have a look at the 'connected hosts' table...but that may not give you what you need, what does though is nmap ...which is available for most all OS'.. including Amiga (https://sourceforge.net/projects/amiga/files/nmap/3.27/ ), so folks can just use that for client discovery... ...what it relies on, is that (most all) onboard NICs are signed by the OEM who made the mainboard ...linux example... Code:
gcb@gallah:~$ sudo nmap -sn 192.168.1.* Starting Nmap 7.70 ( https://nmap.org ) at 2023-02-21 17:35 AEST Nmap scan report for 192.168.1.1 Host is up (0.0095s latency). MAC Address: 34:60:F9:D9:0B:84 (Unknown) Nmap scan report for 192.168.1.50 Host is up (0.012s latency). MAC Address: B8:27:EB:DB:25:BD (Raspberry Pi Foundation) Nmap scan report for 192.168.1.102 Host is up (0.00017s latency). MAC Address: 00:A1:B0:12:86:47 (Unknown) Nmap scan report for 192.168.1.105 Host is up (-0.083s latency). MAC Address: 00:25:22:AB:C9:C2 (ASRock Incorporation) Nmap scan report for 192.168.1.108 Host is up (-0.094s latency). MAC Address: F8:B1:56:A4:B6:D4 (Dell) Nmap scan report for 192.168.1.103 Host is up. Nmap done: 256 IP addresses (6 hosts up) scanned in 4.36 seconds This had me question again, this concept of 'target audience' ~ as I'm firstly aiming at a 'live' deployment on USB stick, I have to imagine some folks might use their own personal computer (and they only have one computer 8), to boot the stick up and play with the emulator ~ how do they get files into the emulator then? USB stick? (doable)...but what if they just want to download/play with demos/games that are in floppy disk format (ADF/DMS)? ![]() All I have to do there is assign the download location to be the fs-uae/floppies dir, and ppl can download, start fs-uae-launcher, and load the disk images =) One could do that in the AmiDeb menu system as well, as Links runs fine on console (bullseye target is links2_2.21-1+b1_amd64.deb), but the debian build of links2 ropes in a bunch of GTK related supports, which might not be useful/wanted there (no GTK stuffs in AmiLFS, I'm not interested in useless 'bloat'} I've moved to booting the UEFI way ~ fungame when you've got a 64bit UEFI image ahead of a 32bit OS but it works. Now I'm organizing the automation, just in case anyone wants to build this thing for themselves ; this happens in the jhalfs interface... ![]() ...that, is a long road, but the system build map is done (complete for intel i915, nvidia not so much =) so it's easy for folks to follow that recipe manually, but not nearly everyone has the inclination/time to do such a thing....it's easier to tell the system to build itself, from a human effort perspective. IMHO everyone into linux should do $LFS just once in their life, to truly appreciate the involved job major linux distro packagers go through all of the time 8) I'm hopeful one mobo I have to check still runs, as I can do an AMDGPU build against that (very close to AMD/Radeon with GPU either as IGP or discrete). ...it is an oblique firefly overlocker.... [ Show youtube player ] |
![]() |
![]() |
#15 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
//... later ...//
...on this machine, if you need wifi connectivity, that's going to happen via USB, and the moment you need to include non-free firmware here to get things working, is the moment you loose GNU purity ~ you are now a whore to the hardware so to speak B) Only way around this, is to stipify something like "only works with wifi dongles with direct linux kernel support and no non-free firmware needed"...nice idea, not really practical. When I purchased the optiplex, it came with a 'sweetener' in the form of 'free' a USB wifi dongle ~ and as I probably expected, it's chipset is some weird flavour of realtek rtl8188, that isn't directly supported by the current linux kernel stock drivers ~ so I had to use an out-of-band driver from github, which works but taints the kernel. That doesn't matter so much, you get used to bending over for the hardware ... but the rest of the wifi stuff was a doddle ... ...I think if I was on the text console, I'd leap to wicd ncurses wifi configurator, but seeing as the aim is to boot straight into X, and because qt5 is around, the wpa_supplicant tarball can build the 'wpa_gui' interface, which works and is easy to use (it even managed to pop some XDG files in the right place, so ICE-wm automagically added that to the network submenu....which I didn't expect and found most civilized). In one specific regard, I get really pissed about the fact certain parties won't entertain the idea of making a magnanimous gester to the world's computing community, by releasing (at least) the Kickstart 1.3 romfiles 'into the wild'... it just sucks imo (https://www.amigaforever.com/kb/13-122) Regardless of what I do here, I know it's up to the end-user to divine where they stand, and how they obtain, the Kickstart romfile of various versions. Why can't they just be like Linus, and just give it away for free...('coz $$$, that's why)...sigh, the world was a simpler place long ago... It actually turned out to be a good choice to use ice-wm for this ~ I've set it to boot to prompt 'old sckool' style ~ it boots to login prompt, you login, type 'startx' and hit enter -- if you blink, you'll miss it, no joke...I'm not even sure my finger leaves contact with the enter key, and ice-wm/X are in your face. There are only 3(4 now) menu-items/programs to choose from in the menuing (aside from essensial things like prefs, xterm, logout etc), and they are of course fs-uae, fs-uae-AROS and fs-uae-launcher. (note to self; add links2 menuentry)...and that's it =) This'll of course just boot straight to nologin xinit in the live USB case. It's easy to give a linux build, that truly 'retro' flavour, without having to do very much ~ having links2 (lynx is also installed), takes you back to a time before the mighty Mozilla was born, pre $netscape times. Normally speaking, having a text based web browser would be a problem, but in reality for the sake of the machine in front of you and what it's used for, your perhaps first choice of visit would be the Aminet site, which works fine ; one might also like to browse other Amiga related archive/scene sites with adf/dms/lha files you can download...no ad banners, pictures, popups, just straight to the file you need/want for the emulator...(you just pre-configure bookmarks in links2 to help facilitate this adventure). I'm just fussing about with boot config now ~ once that's all set, I can straighten the fs-uae staging out ~ idea at this point, is for the fs-uae menu item to launch whatever 'normal use' Amiga Workbench you want to use -- the menu item fs-uae-AROS does the likewise obvious thing, and the fs-uae-launcher does everything else, including arcade mode... ....proftpd is running in the background (as is sshd) - so best to lock these down at the firewall to $localnet only. Most of the netconf stuff is done via dhcp, but you can lock the firewall down to https and be done with it just about... ...bootsplash is fickle ; you have to bootstrap it from grub...so instead of the grub menu you get the bootsplash, which you hinge on fbcon (or efifb), and relinquish once X pops from an exec call..something like this...I'll figure it out (currently on the RT kernel the 'quiet' switch has it head off into lolly land and not boot, and I'm not real sure why yet =). I want some kind of semblance with Amiga Workbench, when ice-wm pops up...ie; an Amiga wallpaper, push the menu bar to the top, only 1 active desktop....stuff like that... ...and that Dell UEFI bios is freaking close to being a pre-emptive OS in it's own right <grin>... I keep wondering if there's a special key sequence that results in it popping an easter-egg game of space invaders of something... lol... that'd be an idea... 8) ...on the side, I'm doing a build of retroarch and the libretro-PUAE core, just to see how that compares to the fs-uae setup. Retroarch compiled, but gagged on Qt5 for some reason, so I just disabled qt5 widgets in config, and you have to specifically enable libusb/HID or it won't make those modules. The specific puae_libretro.so core built straight out of a github clone, and the makefile sets $cflags to -O3 as matter of course. Considering I didn't have to add anything to the $system build to hit the libretro/retroarch target, it's worth checking out as a possible alternative emulator 'presentation'. Right at the now though, it's pretty much all built, and I just have to finalize the configurata. I should mention the VJ clips I link to on YT, are the work of aussie artist Tom Ellard... I lot of this stuff is not in the wild. Some of it was saved from Tom's old sev-com site/BBS, and some bits are taken from SBSD or The Illustrated Family Doctor. If for a moment you detect there's something Amiga-esque about that stuff, Tom used to use Amigas for various tasks (music & video) back in the Amiga heydays, and a good deal of this stuff is cut-up/sampler, and the things one can do with veejay are limited only by your imagination. ...and now, that funky pervert from des moines... and people with cars for heads ..( warning: loud, decrease vol before play) [ Show youtube player ] Last edited by giantclam; 22 February 2023 at 14:13. |
![]() |
![]() |
#16 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
I've never played with ice-wm before ; this thing's got more configuration options than sox...
![]() ...that be the current wallpaper, auto-hide menubar at top.... ![]() ...that menubar is really customizeable ~ it's got the ability to read in the output of top, you can make your own graphics/widgets, and if one worked at it a bit you could get it close to a workbench menubar... ...I found myself between a rock & a tight place, as ice-wm doesn't directly handle desktop icons --- you can make that happen, but if you do, you lose the RMB menu command. One of the things I found really cool on Amiga, was the advent of 'Magic Menus' ... so I figured I'd just stick with that semantic... ![]() ...the py_sqlite gaff -- entirely my fault. Way back when compiling python3 in $base, I didn't know qt5 was on the path, thus python3 was built without the db and sqlite modules ; this is an easy fix by recompiling same version of python3 and reinstalling, but realistically this all should be done during $base, and the jhalfs scripts will be edited to do that ...this of course meant, I had to go back and recompile fs-uae & friends from 3.1.xx and try again there, which is what's seen here ... ![]() Dry run testing ~ this was a lol moment.... just this morning I was chatting in irc about the floppy drive sound emulation, likening it to an internal combustion engine sound generators that ppl buy and stick in/on they're EV bike/car to make it sound like a 'real' vehicle...lol. As a hardware guy, I used to freak out about how some of the disk demos (and some games) didn't treat the floppy drive hardware very kindly, and it used to amaze me the demo could run smoothly and load assets as it thrashed the head carriage all over the disk media seeking load-points. I'll probably forgive my attitude if the disk sounds replicate ultra violence wrt disk accesses when one of these disk images are loaded, or maybe I'd like a Workbench loadup to sound like my old 50MB seagate scsi coffee grinder in the A590 sidecar... ...anyhow, at the local console & startx, then launch fs-uae (no params) to make sure it finds everything, hooks the sound, and initializes the display without segfaulting... and it does... \o/ ...and like, I'm standing there and the little Dell thin-client box starts floppydrive clicking...lol.. and I didn't even have audio jacked in at the back, and it seems I didn't even notice it had it's own internal speaker ... I'm still having a giggle... it's not mentioned in the specs...then it dawned on me ~ this is another feature customers could select available for the businesss/cloud computing 9020 models, that wasn't available for the domestic market machines (which are the only spec sheet available)... maybe a little audio power amp and another decent sounding wee speaker... lol ..dayton transducers?.. ...then things got weird, and remember this has all built and worked fine on Pilot box ~ fs-use was running, but it was in blackscreen. I pulled fs-uae 3.1.66 out of the way, and stuck fs-uae 4.0.x in it's place -- same story, just later in time..hmm. Then I remembered retroarch & libretro-puae core had actually built, so I started that lot up...and it's puae core engine draws to display fine...go figure =) I'll have to figure out why that is so, but from what I've read on the tomes this is likely something to do with mesalib&i915 ...(Pilot is on nouveau), it's the only real difference between the builds, and these HD4600 intel IGPs are batguano crazy (this one initializes 4 dri devices before finding the vga: port for instance}...and all I can think of is puae is using a different GL method?...I'll have to dig the code a bit.. Turns out disabling qt widgets in retroarch (to get it to build) ends up knackering the GUI arrow buttons and suchlike, but you can still drive it with kbd&rodent... it saw the USB Dpad controller, assigned it a number and declared it unconfigured...that sounds about right. Getting the qt widgets to compile, is when you have to presume your LFS build is a long way from what the developers are using...ie; too close to the bleeding edge. I think you shouldn't pester these folks, or raise any tickets, because it's reallt not their problem...ie; when you develop software for distros, you aim for $stable, not $run-with-scissors ... iirc it looked something like gcc-10 prolly gets right, but gcc-12 refuses to parse... same old same old there.. ...and now some words from Dr. Cherry...... [ Show youtube player ] |
![]() |
![]() |
#17 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
Turns out the 9020 USFF is fussy about which mSATA mini_pcie card you stick in it, and it doesn't like the one I bought =^/ I'll have to hunt around for the right type, but that'll be small task ; mSATA is a 10yo standard and the world has moved on (m.2)...and getting one in 2030 size is worse (at least, they're thin on the ground here in AU). The mini_pcie slot was intended for mSATA/wifi card (there's a covered port at the back for wifi antenna), and there was (once upon a time) a caddy unit that replaced the dvd drive with an additional SATA ssd bay, but these seem rare as well (and optical drive may be useful for all the Amiga magazine 'cover' CDs I have in storage)...but if you get a 9020USFF, just know Samsung PM871 msata ssd are known to work (this is what Dell fitted on request)...and everything else is YMMV...
...still trying to figure out why fs-uae is having this "[FSE] [VID] Not a valid drawable size for glViewport" boohiss ... and I still have to look at why puae does not... I ?think? qt widgets in retroarch don't like qt5.15.5 , so I'll rollback to qt5.15.2 as there's not too many things linked against it... (this is fixed in retroarch, was related to which version of libopenssl qt5 was linking against, not qt5 version).. ...it's actually a handy thing having fs-uae-launcher outside the fs-uae tree ~ last time I checked any of this stuff out was around '08 when I was doing some wine-dev stuffs even though wine isn't an emulator. Before that, a decade prior, it would've been E-UAE on ix86 machines which I just recently proved to myself gave the lack luster performance of not enough cpu raw powah... now, it's a different story, but you're still sitting on ix86 ... /// Feb 26th /// Got bitten by a red belly black snake saturday around lunchtime, and spent the next 24hours in hospital. I don't think there's any recorded deaths from bites this aussie snake gives, but seeing as I'm nigh on 65yo with less than half a working liver left, one can't be too careful...I'm fine BTW 8) ...so you can't really look at Retroarch without checking out the 'Lakka' deployment images. It's also interesting/useful to look at how they manage the live USB stick bootup. I tend to approach these things 'as a dummy', and at that level it didn't take me long to get frustrated with the gamepad mappings (perhaps a degree of RTFM is required =), but the keyboard did as it should. Retroarch does a good job with the UI and emulator engine supports, but it does hit you in the face with an "I am an Amiga emulator" ...it's more a case of "I will attempt to emulate every older gaming platform out there"... which in itself is quite a feat. I'll have to boot it up again later... check sound/video actually works on the 9020... I'll probably include it as a menu item, because for mine...seeing x86 emulate Amiga, in turn leads to 'what else can it emulate?' ; Retroarch goes close to answering that question imho. I think I'll polish that Retroarch install off tonight, and just check it's all sane with libretro-puae, then I can circle back and find out wtf fs-uae is complaining about... ...we choose moon ... I hope you see somewhere you know... [ Show youtube player ] |
![]() |
![]() |
#18 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
Amiga owners/users find it relatively easy to get 'close & personal' with their Amiga computer ~ that was one of it's hidden features imho. As a technician, it wasn't just that all the parts of the chipset(s) had familiar human names, or that you found fred & wilma inside your A590 when changing out a long dead 20MB hdrive... it was the OS, the roms, the things the OS could do (that others couldn't) ~ the screen overlays, Agnus functions, and being able to setup RAMB0 and copy workbench floppy to it and warmboot straight to business off the ramdisk...that was/still is... a remarkable thing for the mid 80's (you can do it in linux ofc, but it doesn't really 'present itself' as a user-space device like ramb0 did =), and I can go on and on and on here...and on, and on..
...if you've been into Amigas from day dot (A1000) like myself, then I can hold the above para in analogy, and surely say that doing the $LFS thing is about as 'close & personal' as you're ever going to get with ix86(_64) ... end of story 8) I like to aim for the 'reverse warm fuzzy', wherein you achieve a linux instance that crashes, burns or segfaults on bootup, if it's not on the hardware it was built/compiled for... lol .. ie; will what I've just built on the optiplex 9020, boot and run on the asrock n86-s3?...pffft, nah...kernel's sure to fly into a panic when it can't find $rootfs 'coz there's no mcp-61 driver supports built in...this isn't a 'distro' ; this'd be more something that you'd release specifically targeting the Dell 9020USFF in it's various guises. The replacement mSATA drive arrived this morning ~ 128GB 'GUDGA' brand made in china AU$20 ...clips straight in, seems to work fine ~ if you get/have a 9020USFF I can vouch these work. Personally would've preferred a blue LED. Do I trust it?.... of course not, you can't trust any of this stuff from the get go, but in this specific use case + later configuration to minimize logfile chatter in /var, it'll enjoy a easier life than most SSD drivers (unless I manage to thrash it to death compiling 8) ![]() The obvious pun would be it was bugging me why fs-uae was refusing point blank on the HD4600, while at the same time libretro-puae wasn't quite that refusing, but was still getting the glxviewport slightly wrong. I concluded to build xsceensaver will all plugins, which is where I discovered some of those weren't behaving either..but most others linked to GL & friends worked fine, not all .... but, they ALL worked on the amd/nvidia rig.... ...Occam's razor ~ the $base build had somehow 'tuned' itself to the amd/nvidia rig ; the $base system was not as 'portable' as it first seemed ; I got hoisted by my own words previously typed ...you have to watch out for things that compile and profile themselves against the CPU they find, regardless of typical CFLAGS to persuade the compiler otherwise ... 'gmp' would have been one, libffi is another as well. ( FFI stands for Foreign Function Interface. An FFI allows a program written in one language to call a program written in another language. Specifically, Libffi can provide a bridge between an interpreter like Perl, or Python, and shared library subroutines written in C, or C++ ) ...one if not both of these locked in on the AMD cpu is my hunch (in the same way I saw it yesterday lock in on the Haswell core it identified in the 9020). This is to say, the only way to prove a hypothesis, is to rebuild it all from scratch, on the optiplex itself ($base is done, just moving it to mSATA so I can close the box one last time (hopefully =) ..change of pace ~ 'Matt Frequencies' is an aussie I follow on YT, for his tellings of the BOFH, and vocal acapellas of interesting original material. Of course, he has an Amiga heritage, which often shows in his works ... [ Show youtube player ] Last edited by giantclam; 04 March 2023 at 15:21. |
![]() |
![]() |
#19 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
...."O frabjous day! Callooh! Callay!” He chortled in his joy... \o/ ...
...the thing actually works, I think I'll name it Zebede (Magic Roundabout) ... ..I'd be pretty sure my screwup with gmp/libffi was the crux issue...not that the first ones were screwedup, I'd just created amdfanbois out of them without realizing I'd actually done that ...(lol ... sortta sounds like real-life for intel a lot of the time lately =) ..I found the bios easter-egg ..on the Dell it takes the form of a full GUI automated self test program, but it looks like it can do arcadia/breakout =) In the time between when I started this (LFS 11.2-314), and the now a couple of weeks later with LFS 11.3 ... the world has changed. As in, things that compiled last week no longer do, as the build set components get bumped a version or three closer to the edge. Grumble grumble...python3 is a snake, and I should know having just been bitten by one IRL.. python-3.11.x has some different ideas of how to choke you to death, compared to python-3.10.x .. and my heart goes out to all you package maintainers out there for various linux distros, who have to deal with this ever mobile scene on a continuing basis... ...kudos especially to the debian sid team for the sip-4.19.25-pyframe_getback.patch, and to some fedora python folks for pointing out the 'fix' for some obscure python3 failures was to grab Cython from github and build/install that... both these things saved me much swearing..<grin>.. It's a bit "Oh, Mr Hart!" in /usr/local/src so best I clean that up before somebody notices (the kernel 8)... speaking of which... ...you get a choice of mill (kernel)...it's running a stable release kernel now, I'll have to install Linus' RT tree I was using before, but there's a couple of other kernel trees out for audio/video specific tasks that may be looking into. Or not, considering the optiplex hardware, but it's easy to suck & see. I have to edit ice-wm menu.cfg to remove the 'dangling' "Programs" item, because this linux instance is an Amiga emulator, and there are no 'Programs' as such. Oh and don't forget to fix the sound module loading, and add a menu item for links2 (built with some eye-candy this time)... and as Eric_O from SMA says, "If I can do it, you can do it"... ...oh, the red belly black snake bite? ..mate, they might not be lethal, but that venom sure packs a punch. It's now what, nearly a week later, and my right index finger has swollen up, gone all hard and hot, and has these insatiable itching attacks as akin the the bone of the finger wanting to be scratched ... ever had a plaster cast itch?...worse than that =) Not as bad as a jumper ant sting tho '... one of those got me 2nd toe, and within an hour it was the size of my big toe beside it... itched like mad, took 6weeks for my toe to recover .. lesson: give jumper ants right of way B) ... does your fridge have a problem?... [ Show youtube player ] |
![]() |
![]() |
#20 |
Registered User
Join Date: Jan 2015
Location: australia
Posts: 585
|
Received a AU$15 refund on the machine because SSD wasn't correctly installed, and heatsink transfer compound hadn't been replaced. I'm happy with that, as it almost covers the cost of buying the mSATA drive for the unit... but take that advice onboard ; if the seller's ad says 'professionally refurbished', check and ensure that service has been provided (leaving dried out 10+ year old heatsink compound is never going to be ideal =)
Now that it's all built and running correctly (manually), next trick is automating the bootup/runtime environment ~ for mine, this starts at kernel boot ; I do not want to see kernel messages spray to screen and print-through the boot-logo image ; I don't want ipv6 (slows the bootup), and I don't want sysvinit spraying messages to console either. There are various ways to achieve this ; I choose to do the kernel 'quietening' in the kernel build ; the sysvinit 'quietening' is done in rc.d/bootscripts.... ..right about now, I'd like to bring the little box in here, because after all the effort/time, it finally starts becoming 'fun' -- the thing can be used for intended purpose ; I can start playing with the fs-uae emulation. The fs-uae engine is utilizing a more or less constant 115% cpu time, so the wrapper for tasket has to pin the fs-uae PID to 2 cpu cores. X and ice-wm are <1% cpu time. TBH, I think it's the first time I've had any regard for intel HD46xx IGP ..that said, I'll have to install the intel-igp-toolset tarball, to figure how much of that 115% total is on CPU, and what IGP is utilizing. I had some kind of stoush with fs-uae-launcher (git clone), versus pyqt5 versus python-3.11.2 ...ended up rolling back to fs-uae-launcher 3.2.11-beta (with fs-uae 4.0.40), and everyone's happy again. In fact it all works together very well. I think, end of the day, if you want to get full use out of the emulator, you're gunna need spend money buying an AmigaForever set. It is what it is ~ if you want legit kickstart ROM and Workbench installset, that's what you have/need to do. This is to say, one possible bootup sequence for the emulator, imho, should be power-on -> bootsplash --> Amiga Workbench. You can do that either with or without starting ice-wm. (it probably makes sense to start ice-wm, so if you quit the workbench emulation, you fall thru to X desktop and can access fs-uae-launcher). It's an interesting thought exercise ~ the rhetorical question being, how did/do we quit workbench on Amiga? ...m'kay ... so the closest the emulator can get to that, is by intercepting the APCI event on the 9020USFF power button, and shutting the machine down ; that's one way to 'quit workbench'. The other way, would be Alt-F4 keycombo..this kills the running fs-uae process (and workbench with it)...and that could go on to shutdown the machine, startup an ice-wm session to fall through to...or restart(reboot) the workbench emulation. You could have some faux device-file in workbench/devs called white-rabbit-object, and get the linux init scripts to either check for/parse that file, to either boot to workbench mode, or ice-wm/fs-uae-launcher mode..yeah? (drag white-rabbit-object into workbench/storage to turn emulator workbench bootup off =) I'll have to buy a AmigaForever set for myself and ensure it all rolls into the emulator environment with a minimum of fuss ...in the meanwhile, I'll use my old AIAB installset I bought yonks ago. I'm thinking about AROS (m68k) ... like, why do that when you can use x86_64 AROS instead? I dunno, I suppose it's because I see *uae engines all wearing the AROS romset... ....it reminds me of a time around 2010 (when I was doin' wine-dev), when the windows dosbox/winuae builds were leaps and bounds ahead of the linux/mac ports of the same things, and it turned out viable/worthwhile to run the win32 versions in wine on linux/mac ..ie; the measure of difference between AROS x86 native, versus AROS m68k running is fs-uae. I think that sounds 'pointless' in many respects ~ for mine, if you wanted to do AROS on this machine, you'd do it native from a grub bootmenu option (more on this later). ....and remember, it's all about fresh at hungry jacks... [ Show youtube player ] |
![]() |
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
Thread Tools | |
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
Announcing AmiDeb - An Amiga Emulator system. | vk3heg | News | 44 | 24 July 2024 02:54 |
Amideb: An Amiga Emulator System | giantclam | support.Other | 62 | 16 February 2023 10:18 |
Amilator (amideb) CRASH TEST PULSE AUDIO .log VIDEO | White | support.FS-UAE | 1 | 07 February 2023 04:14 |
Native build system | Edders | Coders. C/C++ | 2 | 30 July 2019 12:18 |
ClassicWB 3.9....new system build-up on WinUAE | NovaCoder | project.ClassicWB | 3 | 15 March 2011 00:42 |
|
|